Several hypotheses have been proposed to explain the maintenance of tropical forest diversity. Three frequently investigated hypotheses focus on the seed and seedling stages when plants are most vulnerable to environmental factors. These hypotheses propose that either niche differentiation, negative density dependence, and/or neutral process maintain biodiversity in tropical forests. The niche differentiation hypothesis proposes that plant species are specialized to microhabitats, as evidenced by differential performance (germination, growth, and or survival) of each species. The second, also termed the Janzen-Connell hypothesis, posits that negative density dependence (i.e., higher pathogen- and predator-induced mortality near conspecifics) regulates the density of common species. The neutral theory maintains that stochastic factors and limited seed dispersal contribute to avoidance of competitive interactions by functionally equivalent species. To investigate these hypotheses in low-diversity tropical forest, I measured seed/seedling dynamics and microhabitats (understory irradiance and substrate) in 4-ha plots in Hawaiian wet and dry forests in which all adult trees were mapped. I found evidence in support of all three hypotheses. Overall, recruitment limitation was the strongest driver of seedling dynamics in Hawaiian wet forest. Recruitment limitations and habitat specialization varied more among species within Hawaiian wet forest than among forests with comparable data. In Hawaiian wet forest, I also found evidence of differential performance among species across microhabitats and striking differences in allometric relationships, suggesting the existence of niche differentiation, though some species-pairs appeared to be functionally equivalent and there was substantial niche overlap in seedling distribution across microhabitats. In both wet and dry Hawaiian forest, density dependence was largely positive, thus it does not appear to maintain coexistence. Altogether, these results show that Hawaiian forest recruitment patterns are complex and are more similar than expected to mainland tropical forests. The results of this study will be useful for identifying and predicting the effects of factors that may be important for tree recruitment at the seedling stage and how these factors vary across species and forest types.